University of Twente Student Theses

Login

Generating physics-based 1D surrogate models from 2D model results

Awakimjan, Ilia (2017) Generating physics-based 1D surrogate models from 2D model results.

[img]
Preview
PDF
10MB
Abstract:Numerical simulations are an essential part of a river engineer’s toolbox. Models are used to predict flow and sediment transport and to solve complex problems. Depending on the problem different types of models are used, often grouped by their dimensionality. Two- dimensional depth-averaged models are used to solve for complex flow conditions. They are sufficiently accurate for intervention design or local flow studies, but are computationally too expensive for certain tasks such as operational forecasting or uncertainty analysis. Increasing the computational speed of 2D models can be done with surrogate modelling. Physics-based 1D surrogate models can be constructed to emulate 2D model results. The aim of such surrogates is to increase computational efficiency while maintaining the accuracy of the 2D model results. One such method has been proposed by Berends et al. (2016). Termed FM2PROF, the method aggregates flow information from 2D model output to construct 1D models. In this thesis cases of various complexity were set up to test the performance of FM2PROF. Based on the initial results several improvements were identified and implemented. The FM2PROF output was verified and the constructed, uncalibrated 1D models were validated. It was found that the method passes verification for simple cases. The produced 1D model cross-section profiles and roughness values are accurate and volume errors between 1D and 2D are small. The validation for the simple cases with uniform profiles shows that the generated 1D models emulate the general 2D model behaviour. The waterlevel errors are between 0 and 25 centimetres, with the 1D model often underestimating the waterlevels. For the complex case of a 2D model of the river Waal, the storage in the cross-section is underestimated, while the main channel roughness section width is greatly overestimated. The 1D model results are also less accurate: for most of the simulation the waterlevels are underestimated up to 1 metre, but overestimations occur of up to 3 metres for low waterdepths. In conclusion it has been shown that the proposed method has potential. For simple cases, physics-based and validated 1D hydraulic river models have been constructed that approximate 2D model results. The validation has also shown that for complex 2D models the constructed surrogates do not accurately reproduce the waterlevels, meaning that either further improvement to FM2PROF or calibration of the constructed 1D models is necessary.
Item Type:Essay (Master)
Faculty:ET: Engineering Technology
Subject:56 civil engineering
Programme:Civil Engineering and Management MSc (60026)
Link to this item:http://purl.utwente.nl/essays/73796
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page